Aggregating smart service duration across process instances

Question

All, 
 I'm looking for a mechanism to report on a specific smart service's execution duration across multiple process instances of a given process model. Some background on my use case: 
 We're doing some performance testing and accessing transaction times is generally trivial via JMeter reporting, however in this particular case, the relevant process runs an asynchronous flow that we're not able to measure directly. As such, I'd like to aggregate node duration within the context of this particular process model so we'll be able to accurately report on the performance of this flow's components. 
 Thanks in advance!

Mike Cichy · Accepted Answer

There currently is no option to get aggregate data on unattended node performance. You could potentially save the performance data in PVs or to a DB for reporting but that could obviously skew your load test results. 
As a proxy metric, you could take execution duration of the process instance that the node is executed in, to estimate the node performance. 
If the node is expected to be slow (over 2s), Health Check will flag it and report on performance based on the existing process instances. You could also try extracting that data from the log that HC creates if you wanted to be more granular. 
If the ss is unique or a plugin, you could get performance data out of smart_service_details.csv if a generic script task then you will not be able to separate the perf data.